I was considering Afshar's experiment and thought of a possible modification, and I was wondering whether it had been proposed, tried, or thought of. If not I was considering suggesting it to Afshar or posting a paper on it. The idea is as follows: Each member of an entangled photon pair is sent either to a lens and detector (signal photon) or a quantum eraser (idler photon). When the idler photon is observed with “which-slit” information intact, the rate of detection of the signal photon is observed in the presence of an Afshar-style interference grid, placed before the lens in the path of the signal photon. It is assumed the grid will have no effect on the detection rate, even when the idler photons are detected with “which-slit” information intact; the only way this can be possible, however, is if the signal photons exhibit interference ([EDIT]wave)-like effects, thereby avoiding the grid via the interference minima, until the instant of detection, at which point their distribution becomes Gaussian. Similarly, for idler photons for which “which-slit” information has been erased, the signal photons are expected to exhibit an interference pattern at all times. The configuration is essentially that of the Delayed Choice Quantum Eraser, except with the Afshar-style grid placed in front of the lens in the path of the signal photon. The experiment can result in two possible outcomes. A. Detections at D0 corresponding to detections at D3 or D4 (which-slit information intact) are significantly fewer in number than detections at D0 corresponding to detections at either D1 or D2 (which-slit information erased). This would only be true if the signal photons corresponding to the detections at D3 or D4 actually traveled a well-defined path between their source and D0, and this path included the possibility of being blocked by the Afshar grid, i.e., they exhibited no self-interference effects. Since the Copenhagen interpretation states the photons do not have a well-defined path prior to detection, this result would be inconsistent with that aspect of the interpretation, but would preserve the notion of complimentarity, at least insofar as a single entangled photon pair never exhibits both wave and particle-like features simultaneously. B. Detections at D0 corresponding to detections at D3 or D4 are just as frequent as those corresponding to detections at either D1 or D2. This is the result expected in the absence of the Afshar grid. This would be true only if the signal photons corresponding to detections at either D3 or D4, with which-slit information intact, nonetheless do not have a well-defined path before detection by D0 and, instead, exhibit self-interference until the moment of detection, allowing them to avoid the Afshar grid, yet still ultimately assemble in a Gaussian pattern. While this preserves the idea that the photons possessed no well-defined path prior to detection, it also would appear to defeat the uncertainty principle, as an unmistakable interference pattern would be (at least) inferable while, simultaneously, which-slit information is obtainable through the joint detection at D0 and D3 or D4. Furthermore, it would be exceedingly difficult to explain this result in any local or realistic model of quantum theory, even in the absence of Bell’s inequality, as the distribution pattern of photons is truly not determined for either member of the entangled pair until both have been detected. Nevertheless, the “no-signaling” condition would suggest that scenario B must be the correct outcome; the presence of the Afshar grid should have no effect on the results. If it did, it would be possible to communicate with a person observing D0 simply by forcing, or not, the erasure of which-slit information. Forced erasure would result in more detections at D0 than when which-slit information is not erased; no coincidence circuit would be required to assemble the results, as the “message” can be reconstructed merely by observing the statistical frequency of detections at D0.